Fixes #19956: Prevent duplicate deletion records from cascading deletions

netbox/core/signals.py

@@ -1,10 +1,12 @@
 import logging
+from threading import local
 
 from django.contrib.contenttypes.models import ContentType
 from django.core.exceptions import ValidationError
 from django.db.models.fields.reverse_related import ManyToManyRel, ManyToOneRel
 from django.db.models.signals import m2m_changed, post_save, pre_delete
 from django.dispatch import receiver, Signal
+from django.core.signals import request_finished
 from django.utils.translation import gettext_lazy as _
 from django_prometheus.models import model_deletes, model_inserts, model_updates
 
@@ -42,6 +44,10 @@
 # Change logging & event handling
 #
 
+# Used to track received signals per object
+_signals_received = local()
+
+
 @receiver((post_save, m2m_changed))
 def handle_changed_object(sender, instance, **kwargs):
     """
@@ -130,6 +136,16 @@ def handle_deleted_object(sender, instance, **kwargs):
     if request is None:
         return
 
+    # Check whether we've already processed a pre_delete signal for this object. (This can
+    # happen e.g. when both a parent object and its child are deleted simultaneously, due
+    # to cascading deletion.)
+    if not hasattr(_signals_received, 'pre_delete'):
+        _signals_received.pre_delete = set()
+    signature = (ContentType.objects.get_for_model(instance), instance.pk)
+    if signature in _signals_received.pre_delete:
+        return
+    _signals_received.pre_delete.add(signature)
+
     # Record an ObjectChange if applicable
     if hasattr(instance, 'to_objectchange'):
         if hasattr(instance, 'snapshot') and not getattr(instance, '_prechange_snapshot', None):
@@ -179,6 +195,14 @@ def handle_deleted_object(sender, instance, **kwargs):
     model_deletes.labels(instance._meta.model_name).inc()
 
 
+@receiver(request_finished)
+def clear_signal_history(sender, **kwargs):
+    """
+    Clear out the signals history once the request is finished.
+    """
+    _signals_received.pre_delete = set()
+
+
 @receiver(clear_events)
 def clear_events_queue(sender, **kwargs):
     """

netbox/core/tests/test_changelog.py

@@ -346,6 +346,38 @@ def test_ordering_genericrelation(self):
         self.assertEqual(changes[1].changed_object_type, ContentType.objects.get_for_model(Interface))
         self.assertEqual(changes[2].changed_object_type, ContentType.objects.get_for_model(Device))
 
+    def test_duplicate_deletions(self):
+        """
+        Check that a cascading deletion event does not generate multiple "deleted" ObjectChange records for
+        the same object.
+        """
+        role1 = DeviceRole(name='Role 1', slug='role-1')
+        role1.save()
+        role2 = DeviceRole(name='Role 2', slug='role-2', parent=role1)
+        role2.save()
+        pk_list = [role1.pk, role2.pk]
+
+        # Delete both objects simultaneously
+        form_data = {
+            'pk': pk_list,
+            'confirm': True,
+            '_confirm': True,
+        }
+        request = {
+            'path': reverse('dcim:devicerole_bulk_delete'),
+            'data': post_data(form_data),
+        }
+        self.add_permissions('dcim.delete_devicerole')
+        self.assertHttpStatus(self.client.post(**request), 302)
+
+        # This should result in exactly one change record per object
+        objectchanges = ObjectChange.objects.filter(
+            changed_object_type=ContentType.objects.get_for_model(DeviceRole),
+            changed_object_id__in=pk_list,
+            action=ObjectChangeActionChoices.ACTION_DELETE
+        )
+        self.assertEqual(objectchanges.count(), 2)
+
 
 class ChangeLogAPITest(APITestCase):